NSF OCI-1147466 : SI2-SSI Collaborative Research: A Computational Materials Data and Design Environment

Abstract: High-throughput ab initio computing has the ability to transform materials research and design, and through its ability to more rapidly invent new materials, invigorate US manufacturing. The time is right to change the paradigm by which materials are designed: Rather than wait for serendipitous discoveries, it is now possible to predict materials properties by solving the basic equations of quantum mechanics, to scale those computations across hundreds of thousands of known compounds, and to explore as many potentially novel materials. The objective of the proposed research is to accelerate materials education, research, and discovery by developing a) the algorithms for high-throughput materials property prediction; b) a database and delivery system for users to access the data, mine it, develop applications from it to determine new properties within large chemical spaces, and dynamically integrate their results with the database for efficient research and development; c) a "computation on demand" capability to easily manage the immensely complex workflows involved in real world materials optimization and new compound design. This Materials Genome Approach: ?? will calculate and make accessible unprecedented amounts of materials data, covering essential properties of the known inorganic compounds. ?? will develop sustainable tools to allow users to rapidly learn properties of a wide-range of material compositions, structures, and properties, including new compounds not existing in nature. ?? will enable users to efficiently predict, screen, and optimize materials at an unparalleled scale. To implement this vision, we have brought together a team with expertise in specific materials domains, computational materials science, and computer science to develop the computational materials data and design environment. The team has helped formulate and initiate many aspects of the Materials Genome concept and therefore bring a unique level of experience and vision to the proposed work. Intellectual Merit: The proposed project provides a significant departure from current materials research methods and is enabled by a sophisticated centralized software architecture system. The system is designed specifically to handle a dynamic continually evolving database, web framework and user interaction. While ab initio computations have already started to show promise for accelerating the traditionally slow development process for new materials, integration with web-based free dissemination and a user-dynamic workspace will lead to a new paradigm for how materials science is performed. In our vision, both experimentalists and theorists will have materials properties of all known inorganic compounds at their fingertips to scan, analyze and provide inspiration for novel materials development. We are envisioning a dynamic 'Google' of materials properties, continually increasing and changing as more users join the community, analyze the data, submit jobs and enter more information back into the framework. Broader Impact: This project will have a wide impact on the materials community by enabling a new approach to materials design. Central to our mission, we are providing free access to our materials design platform, benefitting students in materials science, physics, and engineering, as well as researchers across the world. In addition, we are applying the methodology specifically to the development of better materials for energy harvesting and storage, a problem of critical importance for everyone. This project will also have a broader impact on science, education and society, beyond the field of materials science. The need for storage and web-based dissemination of scientific data is common to all scientific fields and the development of stable, sustainable and accessible frameworks is imperative. We are proposing to make accessible large amounts of data for analysis and data mining but also to allow users to execute their own calculations to impact the data. In this way, we are providing a dynamic scientific design gateway, which could be employed in any computer-enabled science. TPI 7160250

Students:

Weixi Ma, CS, Grad Student
Everett Boyer, Grad Student

Navjeet Sandhu, Grad. Student. UKY - Graduated and has left UKY

Software:

Automated VASP
Pymatgen

Collaborators:

The non-UK collaborators are:
Dane Morgan, Univ. Wisconsin, Prof.
Tam Mayeshiba, Univ. Wisconsin
Gerbrand Ceder, MIT, Prof.
Kristin A. Persson, Research Scientist, LBNL
Anubhav Jain, Research Scientist, LBNL

Title

I am using DLX trying to cast light on a mathematical conjecture concerning Costas arrays (http://en.wikipedia.org/wiki/Costas_array); in particular, I have been trying to find a Costas array of size 32. I was able to show by complete search that if such an array exists, it is not symmetric. The search took about 2000 cpu-days, looking at about 400,000 regions in the search space. I haven't tried yet to relax that condition (which strongly reduces the search space), but I hope to do that. This problem is a good example of a large class of combinatorial problems with search spaces that can be subdivided to arbitrarily small regions for parallel search. I also used the technique to look a bit at Ramsey numbers.

My method involves a coordinating process (written in Perl) with arbitrary many workers (which use a standard answer-set logic solver called clasp). I wrote the coordinating process (complex) and the clasp theory (fairly straightforward). These are available to anyone who is interested.

Software:

Perl (v5.10.1 is installed on DLX) and a download/compilation of clasp (I am using 1.3.10)

Publications

2013

1. Gregory Stump and Raphael Finkel, "Morphological Typology: From Word to Paradigm", Cambridge University Press, 2013.

2012

1. Raphael Finkel and Gregory Stump, "What are principal parts, and what can they tell us about an inflectional system's morphological complexity?" Invited paper presented at the Conference on Morphological Complexity, Convened by the Surrey Morphological Group, British Academy, London, January 13-15, 2012.

2011

1. Raphael Finkel, Barry O'Sullivan, "Reasoning about conditional constraint specification problems and feature models", Artificial Intelligence for Engineering Design, Analysis and Manufacturing 25, pp 163-174, 2011.

2010

1. Raphael Finkel, Heinz Kohler "The trend of amino-acid gain is consistent with a doublet code", Current Proteomics 7, pp 209-211, 2010.
2. Raphael Finkel and Gregory Stump, "Predictability, predictiveness and paradigm complexity". Invited paper presented at the workshop Morphological Complexity: Implications for the Theory of Language, Harvard University, January 22, 2010
3. Raphael Finkel, Gregory Stump, "What your Teacher Told You is True: Latin Verbs Have Four Principal Parts" , in Changing the Center of Gravity ― Transforming Classical Studies Through Cyberinfrastructure, Edited by Melissa Terras and Gregory Crane, pp 287-320, ISBN 978-1-60724-881-1, Gorgias Press, 2010.

Grants

Finkel, Raphael OCI-1147466 SI2-SSI Collaborative Research: A Computational Materials Data and Design Environment $259,070 National Science Foundation 10/1/2012 9/30/2017